With the fast development of computer technologies, there appear to be many information processing devices for accepting users' handwriting input with a handwriting recognition subsystem, such as personal digital assistants PDA or hand portable computers HPC. A handwriting recognition subsystem tends to be useful in the environment of inputting text into a small mobile device like a PDA, or inputting hard-to-enter characters like Chinese into a computer. Users can input handwritten data and symbols into computers by means of pen-like devices. Corresponding to this, there appear to be many handwriting character recognition devices, which can recognize a user's handwriting input.
In the field of handwriting input, two approaches to handwriting character recognition are: on-line character recognition (OLCR) and optical character recognition (OCR). The OCR approach is sometimes also referred to as off-line handwriting recognition. In general, the on-line character recognition (OLCR) technique employs dynamic handwriting information, while the off-line handwriting recognition employs static handwriting information. All OCR systems generally use an input device, such as an optical scanner, for reading text from existing documents into a computer, such as to an image file, and process the image file by data perceiving and data re-construction (e.g. analyze the patterns and identify the characters they represent) to produce a text file for editing or other use later. Relative to the OLCR technique, as the OCR technique cannot obtain real-time dynamic handwriting information such as stroke direction, stroke order, pen tip pressure or speed, etc., as features, the recognition rate will be affected.
The OLCR technique uses a stylus as a handwriting input device to write characters one by one on a digitizing tablet, and then recognizes these characters by a recognition software. In addition to strokes, OLCR technique employs dynamic handwriting information, such as stroke direction, stroke order, tip pressure or speed, etc., as features, it provides generally better recognition accuracy, and is used widely for current handwriting input devices. See the publications entitled, “On-line signature verification by adaptively weighted DP matching,” Authors: Zhao, P. (CADIX Inc., Tokyo, Japan); Higashi, A.; Sato, Y. Source: IEICE Transactions on Information and Systems, Vol. E79-D, No. 5, May 1996, p 535-541; “Recognition of human signatures,” Authors: Pacut, A. (Warsaw Univ. of Technol., Poland); Czajka, A. Source: IJCNN'01. International Joint Conference on Neural Networks, Proceedings (Cat. No. 01CH37222), 2001, pt. 2, p. 1560-4 vol. 2; and “Biometric verification in dynamic writing,” Authors: George, S. E. (Sch. of Comput. & Inf. Sci., Univ. of South Australia, Adelaide, SA, Australia) Source: Proceedings of the SPIE—The International Society for Optical Engineering, v 4738, 2002, p 125-32.
One of the applications of handwriting recognition today is signature recognition for biometric identification and/or verification, most typically used for retail or safeguarding applications, etc. In these applications, the underlying method is to consider the writing pressure (i.e. pen pressure) of a stylus or a pen on a writing surface, in addition to the sequence of x,y-coordinates, as biometric information of a person is considered as a basis for authentication. The authentication technique based on signature verification always utilizes a pressure-sensitive pen and a tablet to record a user's signature. Signature verification then compares the user's signature against a stored signature sample corresponding to the same user, and determines true or false to identify the user.
At present, conventional handwriting input devices utilizing OLCR technique usually request a touch-sensitive pad (e.g. digitizing tablet) which incorporates either magnetic sensor or pressure sensor to sense and record the pen strokes that are touching the pad surface. The conventional digitizing tablet usually has a wire connecting an external smart stylus.
The IBM's ThinkScribe™ is a device integrating a handwriting digitizer having a digitizing tablet with a traditional paper-based recording system. The digitizing tablet includes an active area capable of receiving electromagnetic signals from a radio frequency coupled stylus. This device records a user's handwriting input in strokes and associated timing and can reproduce the user's handwriting input according to the original timing information. A user may write the documents to be transmitted on the digitizing tablet or paper. The digitizing tablet generates a data flow representative of strokes and the associated events, and records the data flow in a nonvolatile memory. The associated events in the data flow may be generally categorized as being either automatically generated by the input device or as being user-invoked. Automatically generated events are events which occur and are detected and recorded without specific input from the user. For example, there may be defined a pen-down event which indicates that the stylus was brought into contact with the writing surface and a pen-up event which indicates that the stylus was lifted from the writing surface. An “ink trace” may thus be defined as a series of pen coordinates recorded between a pen-down and a pen-up event.
All of the input devices mentioned above require a touch-sensitive pad which incorporates either a magnetic sensor or pressure sensor to sense and record the pen strokes that are touching the pad's surface. The pad may be provided at an additional cost to an existing PDA or a personal computer. The pad is also large in size which either is difficult to carry, or it occupies the screen area when it is built onto a PDA and in operation. The pad usually has a wire connecting the pad to the computer, and a wire connecting the pen to the pad. The situation sometimes is a hassle. Besides, considering the identification or verification of the user of a low-cost and low computational power device, such as a mobile phone or a PDA, our choices would then be limited, since it would be unreasonable to attach a high-cost device to collect biometric data for identification or verification, or to adopt a method which needs a lot of computation power to identify or verify the identity of the user. Thus, there is a need to provide a low-cost method and device for signature recognition or authentication which is capable of providing more accurate biometric data (signature data) to a mobile phone or a PDA which only needs reasonable computation power.
In other types of handwriting recognition systems, a pure digital camera input may be used for the sake of handwriting recognition, however, the processing is complicated and the results may not be good. For instance, disclosed in U.S. Pat. No. 6,044,165, assigned to California Institute of Technology, is a technique that uses a digital camera which monitors movement of a writing implement relative to a writing surface, and associated processing hardware which processes the output of the camera to track that movement. However, there are no disclosures to collect biometric data for identification or verification utilizing a digital camera.
As many computer systems (e.g. notebook PC, pervasive device, PDA etc.) are increasingly entering the market equipped with an embedded digital camera of relatively high resolution, it would be advantageous to provide digital video data with the digital camera for use in the handwriting recognition process in such pervasive devices. Accordingly, there is a need to provide an easier low-cost solution to collect biometric data (e.g. signature data) for recognition or authentication which enables a user to write on a paper without a touch-sensitive pad and a wire connecting the sensor to a computer (or a pervasive device and the like) and a wire connecting a stylus (or a pen) to the pad, but equipped with a low-cost digital camera functioning to overcome the known drawbacks mentioned above.